RE: xml:id and opacity of refp's

+1.

In my mind, RefProps/Params go hand-in-hand with the concept of target
Resource which was discussed at length in the WSDL WG many months ago
and was ultimately defeated.

Ugo

> -----Original Message-----
> From: public-ws-addressing-request@w3.org 
> [mailto:public-ws-addressing-request@w3.org] On Behalf Of 
> Savas Parastatidis
> Sent: Thursday, January 06, 2005 3:09 AM
> To: public-ws-addressing@w3.org
> Subject: RE: xml:id and opacity of refp's
> 
> 
> 
> > 
> > 
> > > >  Opaqueness as an operating principle is important to protect 
> > > > service requester interoperability.
> > 
> > Why?  I mean, what's the technical reason for this?
> 
> To my mind, the fundamental issue behind reference 
> properties/parameters and the reason for the value or not of 
> opaqueness is that through RefProps/Params we are creating a 
> placeholder inside an address for service-specific knowledge. 
> This encourages service-builders to couple application-domain 
> specific knowledge with service references (the EPRs). It 
> also moves knowledge from the application level (e.g. 
> resource id, order id, bank account number, etc.) down to the 
> infrastructure (e.g. the way in which we address services). I 
> believe that the two should remain orthogonal but 
> unfortunately RefProperties/Params encourage the opposite, 
> hence encouraging many to treat them as addresses of backend 
> resources.
> 
> That's why I personally prefer WS-Context. It has clear 
> semantics since it clearly identifies a placeholder for 
> defining interaction scope (the semantics are known to all 
> parties involved) but without requiring the semantics of its 
> contents to be known to all. There is no coupling with 
> addresses. The scope of the interaction does not become part 
> of the service reference.
> 
> Please note that the above observations hold true for wsa:action too
> :-))) The purpose of the message payload can be inferred only 
> by the ultimate recipient within the scope of a particular 
> interaction (the interaction being of arbitrary length in 
> numbers of messages).
> 
> Few thoughts to start the 2005 discussions :-)
> 
> Best regards,
> .savas.
> 
> 
>

Received on Thursday, 6 January 2005 15:51:52 UTC